Trying to use dynamic SQL for join column names in SQL Server 2008

Question

Using SQL Server 2008, I would like to specify the column names of an inner join using dynamic SQL. The two tables I am joining have the same names for the columns I am joining on. I know SQL Server does not support natural joins; if it did, the dynamic SQL would look like something like this:

DECLARE @join_columns   AS NVARCHAR(100)
DECLARE @sql_1          AS NVARCHAR(4000)

SET @join_columns = 'Age, Gender' 

SET @sql_1 = '
    SELECT ' + @join_columns + ', table_1.Field_x , table_2.Field_y
    FROM table_1 , table_2
    NATURAL JOIN ON ' + @join_columns

EXECUTE sp_executesql @sql_1

Now, I realize this won't work because there are no natural joins in SQL Server. So, what is the next best way to do this?

Here are a few things I unsuccessfully pursued:

• Tokenizing @join_columns and forming up a dynamic WHERE table_1.<col_1> = table_2.<col_1> [AND...] kind of clause. But, it doesn't look like T-SQL has string tokenization functions.

• Using dynamic SQL to make temp tables, each with a new key column called temp_key that is the concatenation of the fields in @join_column. If it were easy to dynamically concatenate these, then the final join could be always be ON #temp_table_1.temp_key = #temp_table_2.temp_key. One way of setting this up would be to use the REPLACE function to replace the commas in @join_column with plus signs. The problem I ran into here was that the concatenation required casting for the non-VARCHAR columns. So, I'd have to know column types ahead of time - back to square one.

Ideally, I'd like to keep @join_columns as a comma-delimited string, because I am using it elsewhere in dynamic SQL GROUP BY clauses.

It may be that one of the failed approaches above could work, using something I missed. Or, maybe there's a better overall approach.

Any suggestions?

Update

Solution was a combination of both @usr and @Karl's posts below. I used @usr's suggestion to track down a tokenizing table-valued UDF (ended up going with this one ). Then I used @Karl's COALESCE example to turn the resulting table into the WHERE clause. I also used @Karl's full example for another join problem I just ran into. I wish I could give answer status to both posters - thanks guys!

Answer 1

Encapsulate the tokenization/splitting functionality in a table valued UDF. That allows you to build the dynamic SQL string in a clean and architecturally sound way. You can find such splitting functions readily available on the web. It is a shame that they are not built-in but we can build them ourselves.

Answer 2

I find that this works well:

declare @whereClause varchar(8000)
declare @table2 varchar(255)
declare @table1 varchar(255)

set @table1='SomeTable'
set @table2 = 'SomeOtherTable'

SELECT  COLUMN_NAME as [joincolumn]
into    #join_columns
FROM    INFORMATION_SCHEMA.COLUMNS
WHERE   TABLE_NAME = @table1
AND COLUMN_NAME not in ('list names of columns that are not to be joined on here')

select @whereClause=coalesce(@whereClause+' and ','')+ 
'['+@table2+'].'+joincolumn+'=['+@table1+'].'+joincolumn +'
'
from #join_columns

print @whereClause

You can then create a dynamic SQL script and tag the WHERE clause to the back of it